Dioecy in a wind-pollinated herb explained by disruptive selection on sex allocation via inbreeding avoidance

Abstract

The evolution of dioecy from hermaphroditism is widely thought to be a response to disruptive selection favoring males and females, driven by advantages of inbreeding avoidance, sexual specialization, or both. It has hitherto been difficult to uncouple the importance of these two hypotheses. We use phenotypes produced by experimental evolution to test the inbreeding avoidance hypothesis in populations from which sexual specialization can be effectively ruled out. We estimate the selfing rate and the shape of fitness gain curves under scenarios with and without inbreeding depression in experimental populations of wind-pollinated Mercurialis annua with high variation in sex allocation. Our results confirm a phenotypic trade-off between male and female allocation in M. annua. Individual selfing rates increased with pollen production. This dependence led to strong disruptive selection on sex allocation due to its interaction with the mating system under the scenario of high inbreeding depression, especially for plants of medium and large sizes. Taken together, our results demonstrate that inbreeding avoidance on its own can lead to disruptive selection on sex allocation, favoring the selection and maintenance of dioecy under wind pollination without associated benefits of sexual specialization.

Keywords: anemophily; evolutionary branching; geitonogamy; monoecy; paternity; selection gradient; sexual conflict; sexual system.

Fig. 1

Conceptual diagram showing different selection schemes via total fitness on sex allocation (gender) within a population. Here, sex allocation represents how the resource of reproduction is divided into female and male functions in an individual. Unisexual female and male individuals are favored when the selection on sex allocation is disruptive (depicted by the solid line), rendering the population dioecious. By contrast, when the selection on sex allocation is stabilizing (depicted by the dashed line), hermaphroditism is favored.

Fig. 2

The expected effect of plant size on the selfing rates under the mass‐action assumption (see main text for details). Given the same sex allocation (gender), a large plant (shown in blue) will produce more male flowers in absolute terms compared to a small plant (shown in orange). Thus, the expected selfing rate under the mass‐action assumption, which is positively correlated with male flower production, will be higher in the larger plant.

Fig. 3

Plots showing the distribution of gender phenotypes in the studied populations (a), and the trade‐off between female and male flower numbers of the individuals (b), of Mercurialis annua. In (a), a value of zero in gender indicates an individual with only female flowers, whereas a value of one indicates an individual with only male flowers. In (b), the trade‐off lines for plants of different sizes were estimated by the generalized linear mixed model, with the shaded ribbons indicating the 95% confidence interval of the corresponding regression lines. Hereafter, to present the interactive effect with plant size, the regression lines for three levels of size (small, medium, and large) are shown, reflecting plants of mean size minus SD, mean size, and mean size plus SD, respectively. Note that one raw data point with an extreme number of female flowers (2108 female flowers) is not shown to avoid compression of the y‐axis (see also Supporting Information Fig. S1).

Fig. 4

Plot showing the effect of gender (maleness) on the selfing rate of studied individuals of Mercurialis annua (N = 172; individuals producing no mature seed were not included). Note that although the interaction with plant size was not significant, the slope trended to be steeper in larger plants. The shaded ribbon indicates the 95% confidence interval of the regression lines. Note that there was one individual with a gender of zero that had a selfing rate > 0. This inference is likely the result of sampling error during phenotyping, where we might have overlooked male flowers on that individual.

Fig. 5

Plots showing the interactive effects of plant size, degree of inbreeding depression, and gender on relative female (a), male (b), and total (c) fitness of Mercurialis annua. Fitness was estimated under two scenarios of inbreeding depression (δ) of zero and one, depicted by orange and green lines, respectively. Shaded ribbons indicate the 95% confidence interval of the corresponding regression lines. The interactions with population were not significant; thus, the interactive effects with population are not shown (see Table 1, Supporting Information Table S4 for the P values).